Modeling Log S

Abstract

Assessing the aqueous solubility of chemical compounds is one of the main tasks of molecular chemistry, with important implications for drug discovery and other research. In this respect, this study proposes a predictive model for Log S and offers a comprehensive comparative analysis of several machine learning algorithms on six different datasets, evaluating performance in terms of R-squared value, mean square error and execution time on various datasets that differ in composition and source and are used in natural and synthetic contexts. In addition, a performance comparison of the algorithms has been made in order to identify the most efficient one for solubility prediction using molecular data. Results are clearly articulated and presented, with appropriate reasoning and evidence provided. The research question is appropriately addressed in the study by clarifying the differences in the response of the algorithms across six datasets of different compositions and sources. It is clear that the results can facilitate the process of comparing different machine learning algorithms.